BADDE.ORG
DEEN
◀ BACK TO THE HALL
badde.org · Project corridor: truefails

Truefails.com — When AIs Cross-Check Each Other for Truth

A single language model sometimes hallucinates convincingly. Several models that check each other catch much of that. Truefails.com does exactly this via API: statement in, structured truth report out.

The principle: cross-examination, not a single opinion

A claim is sent through several AIs in parallel. Where they agree, confidence rises; where they conflict, the conflict is surfaced with reasoning. So models correct each other based on their respective knowledge.

Strengths and weaknesses, diagnosed

Truefails diagnoses not only the result but the source of uncertainty: outdated knowledge, missing context or a real factual gap. That makes the report usable instead of just a green or red checkmark.

Into any workflow via API

Paste text, upload a CSV or call the API — the report arrives in seconds. Ideal for newsrooms, content pipelines and anywhere AI output must be checked before publishing.

At a glance

  • Several AIs check the same statement
  • Conflicts are surfaced, not glossed over
  • Diagnosis of strengths & weaknesses per source
  • Text · CSV · API

FAQ

How does Truefails handle conflicting AI answers?
Conflicts are not hidden but reported as a conflict with reasoning — you see which model claims what and why.
Can I integrate Truefails into my own pipeline?
Yes, via the API. You send text and receive a structured truth report.
Read more on the project Try Truefails.com